Emergent Cognitive Convergence via Implementation: Structured Cognitive Loop Reflecting Four Theories of Mind

We report a structural convergence among four influential theories of mind: Kahneman dual-system theory, Friston predictive processing, Minsky society of mind, and Clark extended mind, emerging unintentionally within a practical AI architecture known as Agentic Flow. Designed to address limitations of large language models LLMs, Agentic Flow comprises five interlocking modules - Retrieval, Cognition, Control, Action, and Memory - organized into a repeatable cognitive loop. Although originally inspired only by Minsky and Clark, subsequent analysis showed that its structure echoes computational motifs from all four theories. This suggests that theoretical convergence may arise from implementation constraints rather than deliberate synthesis. In controlled evaluations, the structured agent achieved 95.8 percent task success compared to 62.3 percent for baseline LLMs, demonstrating stronger constraint adherence and more reproducible reasoning. We characterize this convergence through a broader descriptive meta-architecture called PEACE, highlighting recurring patterns such as predictive modeling, associative recall, and error-sensitive control. Later formalized as the Structured Cognitive Loop (SCL), this abstraction generalizes principles first realized in Agentic Flow as a foundation for behavioral intelligence in LLM-based agents.Rather than asserting theoretical unification, this position paper proposes that intelligent architectures may evolve toward shared structural patterns shaped by practical demands. Agentic Flow thus functions as an implementation instance of the Structured Cognitive Loop, illustrating how a unified cognitive form can emerge not from abstraction, but from the necessities of real-world reasoning.

💡 Research Summary

The paper reports an unexpected structural convergence between four influential theories of mind—Kahneman’s dual‑system model, Friston’s predictive processing, Minsky’s society of mind, and Clark’s extended mind—and a practical AI architecture called Agentic Flow. Agentic Flow was originally built to mitigate the reasoning and control shortcomings of large language models (LLMs). Its design consists of five interlocking modules—Retrieval, Cognition, Control, Action, and Memory—organized into a repeatable cognitive loop. Although the initial inspiration came only from Minsky and Clark, a retrospective analysis revealed that each module mirrors core mechanisms from all four theories.

The Retrieval module supplies fast, context‑relevant information from external knowledge bases, functioning like Kahneman’s System 1 (rapid, associative processing). The Cognition module, powered by an LLM, generates hypotheses and performs associative reasoning; it also implements Friston’s principle of minimizing prediction error by comparing generated predictions with incoming feedback. The Control module acts as a meta‑cognitive monitor, analogous to Kahneman’s System 2, Minsky’s “monitoring agents,” and Friston’s error‑sensitive hierarchical control. It decides when to intervene, re‑prioritize tasks, or trigger additional reasoning. The Action module executes commands in the environment, while the Memory module stores short‑term and long‑term traces, embodying Clark’s extended‑mind thesis that cognition extends into tools and external artifacts.

The authors abstract this recurring pattern into a meta‑architecture named PEACE (Predictive, Associative, Control, Execution, and Memory) and formalize it as the Structured Cognitive Loop (SCL). SCL generalizes the Agentic Flow design, providing a blueprint for behavioral intelligence in LLM‑based agents.

Empirical evaluation compared Agentic Flow against baseline LLM agents on a suite of multi‑step reasoning and tool‑use tasks. Agentic Flow achieved a 95.8 % task‑success rate, dramatically outperforming the baseline’s 62.3 %. The performance gap was most pronounced in scenarios requiring error detection, plan revision, and integration of external tools—situations where the structured loop’s predictive and control mechanisms provide robustness.

Key insights include: (1) the “fast retrieval‑slow verification” constraint inherent in human cognition emerges naturally in engineered systems and aligns with all four theories; (2) hierarchical error‑sensitive control mirrors predictive processing in the brain; (3) coupling of memory and action with external artifacts validates the extended‑mind perspective; and (4) theoretical integration can arise from implementation constraints rather than deliberate synthesis.

The paper concludes that Agentic Flow serves as a concrete instance of a unified cognitive form that can emerge from practical design pressures. This suggests that future AI architectures may benefit from deliberately incorporating motifs identified across cognitive science, leading to more reliable, interpretable, and human‑like reasoning systems.